Smart Bidding Optimized for $45 First Orders. Each Subscriber Was Worth $360. It Was Finding the Wrong Customers.

The situation

An Australian subscription box brand was spending $28K/month across Meta and Google. Every new subscriber paid $45 for their first box. Smart Bidding was targeting a $40–50 CPA — which appeared profitable. But their data showed the average subscriber stayed 8 months at $45/month: a lifetime value of ~$360. The algorithm had never seen any of that.

The problem

The tracking was standard: one purchase event for the initial $45 order. That was all the platforms knew. From Meta's and Google's perspective, every customer was worth exactly $45. In reality, some subscribers churned after 2 months ($90 LTV) while others stayed 18 months ($810 LTV). Smart Bidding was finding people likely to buy a first box — not people likely to stay subscribed. By optimizing on first purchase alone, they were systematically attracting chronic churn customers and missing the high-LTV cohorts who made the economics work.

What we did

1. Tagged all recurring Shopify subscription billing events with the same customer identifier used at initial signup
2. Configured Google Ads Customer Lifetime Value — each repeat billing event uploaded as a conversion with actual revenue, linked to the original GCLID from acquisition
3. Built a Meta Value Optimization setup — recurring charges uploaded via Offline Conversions API with hashed email and billing amount, attributed to the original Meta ad
4. Created Meta custom audiences by subscription duration (1–3 months, 4–8 months, 9+ months) — lookalikes built exclusively from the 9+ month high-LTV cohort
5. Shifted both campaigns to optimize on LTV — Meta Value Optimization and Google Target ROAS using 6-month cumulative billing revenue

Why it worked

Subscription businesses that optimize on first-purchase CPA are solving the wrong equation. Smart Bidding cannot optimize for LTV if it has never been told what LTV looks like. Once Meta and Google could see that a January subscriber was still paying in September, they fundamentally changed which audience they targeted — and found customers who stayed three times longer.